Bug in sub?

[Chalmers]

Hi,

I assume it’s been a long day and I’m missing something but line 3 below
sure looks like a bug to me:

$ irb
irb(main):001:0> “a”.sub(“a”, “+”)
=> "+"
irb(main):002:0> “a”.sub(“a”, "\")
=> "\"
irb(main):003:0> “a”.sub(“a”, “\+”)
=> ""
irb(main):004:0> exit

robert_feldt@it002473 /tmp/ruby
$ ruby -v
ruby 1.9.0 (2004-04-04) [i386-cygwin]

The same behavior is seen in 1.8.1 i386-cygwin and 1.8.1 i386-mswin32
and for gsub so it’s been in there for some time.

What am I missing?

Regards,

Robert Feldt

[ts1 (ts)]

irb(main):003:0> "a".sub("a", "\\+")
=> ""

svg% ruby -e 'p "abcdef".sub(/(.)./, "\\+")'
"acdef"
svg%

svg% ruby -e 'p "abcdef".sub(/(.).(..)/, "\\+")'
"cdef"
svg%

Guy Decoux

[Chalmers]

ts wrote:

“R” == Robert Feldt feldt@ce.chalmers.se writes:

irb(main):003:0> “a”.sub(“a”, “\+”)
=> “”

svg% ruby -e ‘p “abcdef”.sub(/(.)./, “\+”)’
“acdef”
svg%

svg% ruby -e ‘p “abcdef”.sub(/(.).(…)/, “\+”)’
“cdef”
svg%

Yeah, that’s right the “last matched group”; I don’t use that one much
apparently…

Thanks Guy!

/Robert

[Martin_Hart]

Sorry, call me thick but I don’t understand that

Did Guy confirm that it was a bug, or show that it is not a bug?

Thanks
Martin

···

On Tuesday 06 Apr 2004 3:01 pm, Robert Feldt wrote:

Yeah, that’s right the “last matched group”; I don’t use that one much
apparently…

[Benedikt_Huber1 (Benedikt Huber)]

Perhaps I’m missing the point, but what semantic have consecutive
backslashes substituting a string ?
I feels like a bug, but probably isn’t:
irb(main):008:0> puts “abcd”.sub(‘abcd’,“\”).length
1
irb(main):014:0> puts “abcd”.sub(‘abcd’,“\”*2).length
1
irb(main):009:0> puts “abcd”.sub(‘abcd’,“\”*10).length
5

Thx a lot,
benedikt

···

On Tue, 06 Apr 2004 23:01:55 +0900, Robert Feldt wrote:

ts wrote:
…

svg% ruby -e ‘p “abcdef”.sub(/(.).(…)/, “\+”)’
“cdef”
svg%
…
Yeah, that’s right the “last matched group”; I don’t use that one much
apparently…

[Ara.T.Howard]

not_a_bug # that’s python for not a bug

svg% ruby -e ‘p “abcdef”.sub(/(.)./, “\+”)’
“acdef”
svg%

/(.)./ and ‘+’

- match a char, followed by another char. remember the first char
- replace the entire match with the remembered first char

abcdef -> acdef

···

On Tue, 6 Apr 2004, Martin Hart wrote:

On Tuesday 06 Apr 2004 3:01 pm, Robert Feldt wrote:

Yeah, that’s right the “last matched group”; I don’t use that one much
apparently…

Sorry, call me thick but I don’t understand that

Did Guy confirm that it was a bug, or show that it is not a bug?

--         
-         -

svg% ruby -e ‘p “abcdef”.sub(/(.).(…)/, “\+”)’
“cdef”
svg%

/(.).(…)/ and ‘+’

- match a char, followed by another char. followed by two more chars.
  remember the last two chars (and the first) 
- replace the entire match with the remembered last two chars  (which is
  the 'last matched group')

abcdef -> cdef
----        
  --      --

-a

===============================================================================

EMAIL :: Ara [dot] T [dot] Howard [at] noaa [dot] gov
PHONE :: 303.497.6469
ADDRESS :: E/GC2 325 Broadway, Boulder, CO 80305-3328
URL :: Solar-Terrestrial Physics Data | NCEI
TRY :: for l in ruby perl;do $l -e “print "\x3a\x2d\x29\x0a"”;done
===============================================================================

[Mark_Hubbart]

normal strings are escaped thusly:

puts “\”,“\\”,“\\\”

\
\
=> nil

regexp strings have and extra level of escaping, so that you can
include literal “\1”'s in your substitution. So, they are escaped
thusly:

puts “”.sub(//,“\\”), “”.sub(//,“\\\\”),
“”.sub(//,“\\\\\\”)

\
\
=> nil

The somewhat confusing part is that a backslash in a gsub that doesn’t
translate to a substitution expression becomes a literal. So, since
there is nothing for the last backslash to escape:

puts “”.sub(//,“\”), “”.sub(//,“\\\”), “”.sub(//,“\\\\\”)

\
\
=> nil

… another reason why, as someone pointed out in a thread earlier this
month, it’s handy to only use the block form, and avoid the argument
form of (g)sub like the plague

cheers,
–Mark

···

On May 10, 2004, at 11:38 AM, Benedikt Huber wrote:

On Tue, 06 Apr 2004 23:01:55 +0900, Robert Feldt wrote:

ts wrote:
…

svg% ruby -e ‘p “abcdef”.sub(/(.).(…)/, “\+”)’
“cdef”
svg%
…
Yeah, that’s right the “last matched group”; I don’t use that one much
apparently…
Perhaps I’m missing the point, but what semantic have consecutive
backslashes substituting a string ?
I feels like a bug, but probably isn’t:
irb(main):008:0> puts “abcd”.sub(‘abcd’,“\”).length
1
irb(main):014:0> puts “abcd”.sub(‘abcd’,“\”*2).length
1
irb(main):009:0> puts “abcd”.sub(‘abcd’,“\”*10).length
5

Thx a lot,
benedikt

[Chalmers]

Ara.T.Howard wrote:

Sorry, call me thick but I don’t understand that

Did Guy confirm that it was a bug, or show that it is not a bug?

not_a_bug # that’s python for not a bug

svg% ruby -e ‘p “abcdef”.sub(/(.)./, “\+”)’
“acdef”
svg%

/(.)./ and ‘+’

• match a char, followed by another char. remember the first char
• replace the entire match with the remembered first char

abcdef → acdef

•     -
  

svg% ruby -e ‘p “abcdef”.sub(/(.).(…)/, “\+”)’
“cdef”
svg%

/(.).(…)/ and ‘+’

• match a char, followed by another char. followed by two more chars.
  remember the last two chars (and the first)
• replace the entire match with the remembered last two chars (which is
  the ‘last matched group’)

abcdef → cdef

 --      --

Yes, and further: If you want to do substitutions and aren’t sure there
will be no backslash sequences in the replacement string you should use
the block form, like so:

$ ruby -e ‘p “a”.sub(“a”) {“\+”}’
“\+”

I added a note to my code review checklist to “always” use the block
form…

Sorry for wasting bandwidth on this; I should have read the docs more
closely.

/Robert

[Mark_Hubbart]

only one problem with that:

mark@imac% cat test.rb
require ‘profile’
def gsub_block_test(string, pat, str)
10000.times{ string.gsub(pat){str} }
end
def gsub_arg_test(string, pat, str)
10000.times{ string.gsub(pat,str) }
end

gsub_arg_test(“testing this out”, /[aeiou]/, “.”)
gsub_block_test(“testing this out”, /[aeiou]/, “.”)

mark@imac% ruby test.rb
% cumulative self self total
time seconds seconds calls ms/call ms/call name
54.97 13.65 13.65 20000 0.68 0.68 String#gsub
44.95 24.81 11.16 2 5580.00 12405.00 Integer#times
0.48 24.93 0.12 1 120.00 120.00
Profiler__.start_profile
0.00 24.93 0.00 2 0.00 0.00
Module#method_added
0.00 24.93 0.00 1 0.00 16550.00
Object#gsub_block_test
0.00 24.93 0.00 1 0.00 8260.00
Object#gsub_arg_test
0.00 24.93 0.00 1 0.00 24830.00 #toplevel

the block version of this same test takes twice as long

–Mark

···

On Apr 6, 2004, at 3:29 PM, Robert Feldt wrote:

I added a note to my code review checklist to “always” use the block
form…